The first main challenge for neuroergonomics is to implement an innovative methodology to avoid the pitfalls of reductionist approach leading to place participants facing repetitive and boring artificial tasks causing degradation of data due to attentional and motivational effects as well as investigating task load with realistic/real-world complex cognitive tasks, with studies ranging from pilots, air traffic controllers, surgeons, car drivers, teacher-students in classroom to pedestrians navigating outdoors (Ayaz et al., 2012; Mühl et al., 2014; McKendrick et al., 2016; Arico et al., 2017; Unni et al., 2017; Bevilacqua et al., 2018; Di Flumeri et al., 2018; Gateau et al., 2018; Callan and Dehais, 2019; Djebbara et al., 2019; Modi et al., 2019; Wunderlich and Gramann, 2020). Thus, the challenge for neuroergonomics is to design an engaging ecological paradigm while assuring a high experimental control level. High-quality neuroergonomics research should proceed by conducting a continuum of experiments starting with controlled protocols with high spatial resolution devices (e.g., fMRI, MEG) progressing to more ecological experiments in dynamic micro-worlds using portable devices that are portable but with lower accuracy, to eventually conducting less controlled experiments in a realistic simulated environment and the real world (Ayaz and Dehais, 2019). This implies the use of complex and continuous natural stimuli (e.g., videos, speech, simulator) and the implementation of appropriate signal conditioning and statistical methods to decode the associated brain response on the fly (see, for example, Wong et al., 2018). Much progress has to be made in this direction, including the design of novel protocols and methods to investigate individual differences. This is an important issue as it may affect the sample size, induce biases of the results and the interpretation at the group level, and decrease single-trial classification accuracy. This is also a relevant to the design of personalized products, services, and neuroadaptive technologies and to improve the efficiency of neurostimulation and neurofeedback.

Another challenge for neuroergonomics is to develop and shape its own concepts rather than borrowing constructs from other fields. Following this perspective, efforts should be put toward defining new concepts to assess and predict human performance per se. For example, most of the research to-date has focused on the measurement of mental workload, but this transactional construct remains difficult to operationalize despite decades of research and more than 200,000 publications since the early 2000s'. As a measure, the mental workload is highly sensitive to inter and intra-individual variability and is limited to providing a non-specific and global indicator rather like a thermometer. However, unlike a thermometer, it does not give access to an absolute and reliable assessment of degraded performance. We advocate rather to identify neurophysiological, physiological and behavioral markers that specifically account for degraded mental states ranging from task disengagement (e.g., mind wandering, effort withdrawal) to task over-engagement (e.g., attentional tunneling, perseveration, inattentional blindness, and deafness see Dehais et al., 2020 for a comprehensive review). This approach will allow us to design neuroadaptive technology to monitor the mental states and trigger appropriate cognitive-countermeasures to mitigate their deleterious effects on human performance.

There is also a myriad of methods to measure cerebral activity, and new tools and innovative approaches are published every year. For instance, a recent and exciting trend is to implement connectivity metrics over fMRI, MEG, EEG, or fNIRS signals to identify neural pathways and brain dynamics (Farahani et al., 2019). It offers exciting prospects for hyperscanning purposes to study social cognition at the cortical level. Researchers have a choice between model-free vs. model-based, Granger-based, phase-based, information-based, or high-order and non-linear methods. Each of these methods can be implemented via different formalisms (e.g., phase slope index, phase-locking factor, phase-locking values). There is a critical need to establish a consensus and recommend that authors benchmark these methods when reporting their results rather than publishing only the successful ones.

The rapid evolution of personal electronics in the last quarter century has seen remarkable innovation in computers, digitization of sensors and adoption of wearable sensing technologies. Data that could only be collected in laboratory environments is nowadays more accessible, affordable, and easily integrated into popular electronics and other smart devices. This technological development has moved quickly to meet a growing demand for health-analytics, driven by individuals who wanted to learn more about themselves. While, modern activity trackers have introduced physiological measurements such as heart-rate to measure physical exertion or correlates of stress, these measures are fundamentally non-specific, thereby leaving a picture that is far from complete. In order to expand on the features available currently and elevate the role of continuous monitoring solutions in work and at home, the application of wearable sensors must be expanded to new areas and nowhere is there more untapped potential than in the brain (Curtin and Ayaz, 2018).

New applications are now possible due to wearable and mobile nature of portable neuroimaging like investigating neural correlates of spatial navigation and movement (Djebbara et al., 2019), multi-brain interaction (Liu et al., 2017), practical brain-computer interfaces (Zander et al., 2016) and brain to brain interfaces (Jiang et al., 2019). In fact, portable neuroimaging miniaturization efforts started early 2000s (Ayaz et al., 2013; Yücel et al., 2017) and the latest generation of mobile battery operated and wireless systems have allowed monitoring brain activity and investigating cognition-in-action, in increasingly realistic and real-world settings like participants walking outdoors (McKendrick et al., 2016), students in classroom (Poulsen et al., 2017), air-traffic-controllers working on radars (Ayaz et al., 2012), surgeons performing operations (Shewokis et al., 2017; Modi et al., 2019) to pilots flying aircrafts up in the sky (Gateau et al., 2018).

Technical progress has led to the design of portable wireless devices that allow reasonable degrees of freedom of movement to measure brain activity under increasingly naturalistic settings. The development of dry electrodes expanded rapid out-of-lab applications for EEG and has been shown to provide useable signal quality (Zander et al., 2017). Moreover, for optical brain imaging, fiber-less flat sensors were able to increase the surface area of light detectors to increase resistance for motion artifacts, much needed for mobile applications (Ayaz et al., 2013; Wang et al., 2017). However, there are still obstacles plaguing the widespread use of portable and ultra-portable sensors. Several new directions have to be taken to improve and generalize the use of portable neuroimaging for everyday life situations, and specifically at home settings and in clinical applications. For instance, dry-electrode and fiber-based sensors inevitably lead to significant discomfort for participants after an hour. New mobile solutions such as unobtrusive cEEgrid, in-hear electrodes, miniaturized prefrontal adhesive-optodes could allow to overcome these issues. However, one should be aware that these solutions reduce their potential to account for the complexity of the brain. EEG and fNIRS still face some inherent limitations, respectively, in terms of spatial and temporal resolutions. A recent trend is to perform concurrent fNIRS and EEG recording to overcome each other measurement weaknesses and to give a better insight on neurovascular coupling. Again, the combination of these techniques has a cost in terms of setting time, weight, and discomfort for participants. These settings generally result in a non-ideal compromise in terms of signal quality either for fNIRS or for EEG. Hardware should be developed to fully integrate these sensors in a non-intrusive fashion. Futuristic novel approaches are under development, such as holographic optical brain monitoring re-forms light passing through the body may also revolutionize the scanning of the brain in the real world. Similarly, portable magnetoencephalography portable devices are starting to be developed and becoming a new complementary media to investigate human cognition (Boto et al., 2018).

Non-invasive brain stimulation techniques such as transcranial magnetic stimulation (TMS) and variants of transcranial electrical stimulation (tES) such as transcranial direct current stimulation (tDCS) can alter brain function by positioning the actuator devices over the scalp. Such neuromodulation approaches have led to a proliferation of research on the brain and cognitive augmentation, both in healthy adults and in patients with neurological or psychiatric disease. Neuromodulation has shown a lot of promise in the treatment of heterogeneous psychiatric disorders and TMS is already FDA approved for multiple indications: major depression, migraine pain, and obsessive-compulsive disorder. The tES systems have the potential for portability and, new methodological developments (Knotkova et al., 2019) offer new vistas for research and clinical use and even at-home settings (Charvet et al., 2020). Furthermore, integration of these technologies with neuroimaging provides opportunities for read-write brain-computer interfaces, that can both monitor and alter brain activity, (see Cinel et al., 2019 and Rao, 2019 for descriptions and Jiang et al., 2019, for a closed loop multi-brain example) as well as rehabilitation for diverse clinical indications (Teo et al., 2016). There are already best-practices guidelines and systematic reviews for multimodal use: tDCS+fMRI (Esmaeilpour et al., 2020), TMS+fNIRS (Curtin et al., 2019), TMS+EEG (Darmani and Ziemann, 2019).

Moreover, newer techniques have been emerging. Focused ultrasound (FUS) promises to bring together strong features of TMS and tES, high spatial resolution and wearable mobile footprint, respectively, into a single system. So far it has been studied with animal models, and a few human studies have been reported. There are still significant engineering and safety challenges before effective research systems could be available.

A more futuristic neuromodulation technique is optogenetics: a combination of genetic and optical methods for targeted and fast control of neural activity (Deisseroth, 2011). Optogenetics utilizes light for milli-second resolution speed and cell-type specific precise control for neurons. As it requires viral injection to modify the genetic material of target neurons (to insert light-sensitive receptor protein on the cell membrane), this technique cannot be used on humans and is not currently compatible with neuroergonomic approach. However, it has become a workhorse of neuroscience to understand diverse neural systems both in healthy and disease animal models.

Recent trends in neuroergonomics and neural engineering have used neurotechnologies to enhance various human capabilities (cognitive function like attention, decision-making, mood), and treat neurological and psychiatric disorders (Valero-Cabre et al., 2017; Cinel et al., 2019). Neurotechnologies for brain stimulation are evolving rapidly, and new advances in the techniques and applications are expected in near to mid-term. However, research is still needed to assess the effectiveness of such technology as some studies found that effects remain inconsistent (Bestmann et al., 2015; Horvath et al., 2015), and new guidelines for realistic environments are emerging only very recently (Bikson et al., 2020; Charvet et al., 2020).

Another significant development that will have a substantial impact on future advances in neuroergonomics is the rapid progress in artificial intelligence (AI) (Jason, 2017). Artificial intelligence technology, particularly machine learning (ML), has become ubiquitous for solving many complex problems in our daily lives (Bengio et al., 2013). Many AI approaches, including, for example, Support Vector Machine (SVM) and Fisher Linear Discriminant (FLD), have been successfully used for the assessment of brain states based on fMRI data (Mourao-Miranda et al., 2005). The analysis of fMRI scans by means of SVM has also been applied to a sensory-motor task (Wang et al., 2007). Furthermore, it was demonstrated that the functional brain analysis with neurophysiological interpretation could be facilitated by transforming neural networks “backward models” into “forward models” which is applicable for both EEG and fMRI experimental data (Haufe et al., 2014; Li et al., 2018). Deep learning (DL), a subset of ML, has shown remarkable progress in recent years, including applications to assessment of human performance, with a variety of applications such as image classification, speech recognition, or natural language processing (LeCun et al., 2015; Schmidhuber, 2015). Recently, DL with convolutional neural networks has been successfully applied for EEG decoding and visualization purposes (Schirrmeister et al., 2017). The use of such an approach remains challenging because neuroergonomics experiments generally lead to collect small data samples thus limiting the application of DL techniques. We encourage researchers to share their data with the community following the Brain Imaging Data Structure (BIDS) recommendations so as to constitute very large database to improve our understanding of the brain via AI. Such big data of brain could enable larger scale teamwork across interdisciplinary teams and multi-site collaborations.

Recent advances in the theory and applications of explainable artificial intelligence (XAI) to neuroscience (Fellous et al., 2019), is of high relevance to the field of neuroergonomics. Humans beings are already interacting with robots and AI-based algorithms, and this trend will continue to increase. It is of great importance to ensure transparency and explainability of these artificial decisional systems and their evolution over time to improve trust and optimal human-machine teaming. Currently, the main areas of XAI research in the domain of neuroscience include (1) identifying how explainable learning solutions can be applied, (2) developing a community of scholars working with XAI, and (3) stimulating an open exchange of data and theories. In the near future, it will be possible to apply XAI techniques for intelligent decoding and the modulation of behaviorally activated brain circuits to improve our understanding of human behavior in real-world settings. For example, Searchlight, a sophisticated form of XAI approach, has been developed for the visualization of fMRI, allowing to identify differences in the regional spatial activity pattern of the brain under a variety of experimental conditions (Kriegeskorte et al., 2006). Recently, the potential value of XAI in the field of neurostimulation was demonstrated by Fellous et al. (2019). However, as recently discussed by Arrieta et al. (2020), the field of XAI faces many challenges, including setting the objective metrics on what constitutes a good explanation. For example, Páez (2019) noted that the term “explanation,” as it is currently used in XAI, does not share the properties that are attributed to explanations in epistemology and the philosophy of science, and that the interpretative AI models can provide false assurances of comprehensibility. Furthermore, the barrier of explainability brought by sub-symbolism of deep neural networks initiated the discussion of responsible artificial intelligence, i.e. the ways to implement AI methods with fairness, model explainability, and accountability (Leslie, 2019; Arrieta et al., 2020). Finally, Byrne (2019) pointed out that XAI could benefit from including the rich knowledge in cognitive science about human reasoners' cognitive capacities.

This approach can provide promising prospects for the challenges proposed by the National Academy of Engineering (National Academy of Engineering, 2008) to “reverse engineer the human brain.” The realization of such a challenge requires a much better understanding of brain structure and its cognitive functions, and, among other advantages, would lead to the development of general-purpose artificial intelligence, facilitate human learning, improve methods for diagnosing, treating, and monitoring mental illnesses in a personalized manner, or develop a variety of neuroprosthetics (Roysam et al., 2009). It was also observed that the reverse-engineering the brain is critical to understanding the human mind that will have a profound impact on the future advances in technology, health, and society at large. Clearly, neuroergonomics discipline can contribute to and benefit from the realization of this grand challenge.

The advances of neurotechnologies that can record and alter brain activity provide increasingly powerful tools for neuroergonomists and allied professionals and, are rapidly transforming research with implications for everyday life. It is clear that these more powerful methods mean more responsibilities for neuroergonomists. This calls for a new age “neuroergonomics philosophy” as recently defined by Onaral (2021) and informs rethinking of the “neuroergonomic ethos.”

The field of neuroethics emerged in early 2000s triggered with the awareness of advances in cognitive neuroscience (Farah, 2005). Neuroethics encompasses a large and varied set of issues, from practical considerations, like use of neurotechnology or the data, to more philosophical issues. Several ethical and philosophical issues regarding have been recently discussed (Levy, 2007; Giordano and Gordijn, 2010; Farah, 2015; Amadio et al., 2018; Zuk et al., 2018). It can be argued that neuroergonomics faces similar ethical and philosophical challenges as cognitive neuroscience. For example, Farah (2005) and Farah (2015) discussed the implications of the developments in neurotechnology for individuals and society, including the philosophical concerns, including the impact of cognitive neuroscience on the way we think about ourselves as persons, the issue of moral agents, and spiritual beings, the nature of mind, and, ultimately, how, neuroscience will shape the future of individuals and society at large. Recently, Amadio et al. (2018), discussed several neuroethics questions to guide ethical research concerning the potential impact of a model or neuroscientific account of disease on individuals, communities, and society. For example, among many, the following set of considerations, which are also relevant to the field of neuroergonomics, have been posed:

New ethical challenges that were hypothetical previously are becoming more practical concerns, like who should have access to brain activity patterns, changing personalities with brain stimulation, and mind reading and writing. Considering the growing privacy concerns even about the web surfing behavior, and that social media activity can be used and abused by third parties, the privacy in the age of neuroergonomics is ever vulnerable. Privacy of our brain is the most sensitive data of all as it encapsulates our inner-most thoughts and intents, and the most critical to protect for individuals and society at large. Akin to basic human rights, we should define and protect our neuro rights, fundamental to cognitive freedom and personhood (Ienca and Andorno, 2017).

Responsible neuroergonomics research requires awareness of current neuroethical challenges and continuous interaction among all stakeholders: researchers, ethicists, regulators, lawmakers, lawyers and public at large. As new innovations emerge in neuroergonomic technologies, methods and application, so do new ethical dilemmas. For example, ethical issues specifically arising from the use of portable neuroimaging is recently outlined (Shen et al., 2020), multi-person brain to brain interfacing (Hildt, 2019) and, benefits/risks of neuroenhancement of surgeons (Patel et al., 2020). Recently, Farahany and Ramos (2020) highlighted that neuroethics can foster necessary and beneficial collaborations for responsible neuroscientific discovery. Neuroethics should be an integral part of neuroergonomics, and with the use of foresight, open communication, strategic planning, we can all truly benefit from the advances of neuroergonomics research and development.